Download Optimal EPI parameters for reduction of susceptibility

www.elsevier.com/locate/ynimg
NeuroImage 33 (2006) 493 – 504
Optimal EPI parameters for reduction of susceptibility-induced
BOLD sensitivity losses: A whole-brain analysis at 3 T and 1.5 T
Nikolaus Weiskopf ,⁎ Chloe Hutton, Oliver Josephs, and Ralf Deichmann
Wellcome Department of Imaging Neuroscience, Institute of Neurology, University College London, London, WC1N 3BG, UK
Received 24 April 2006; revised 16 July 2006; accepted 18 July 2006
Available online 7 September 2006
Most functional magnetic resonance imaging (fMRI) studies record the
blood oxygen level-dependent (BOLD) signal using fast gradient-echo
echo-planar imaging (GE EPI). However, GE EPI can suffer from
substantial signal dropout caused by inhomogeneities in the static
magnetic field. These field inhomogeneities occur near air/tissue
interfaces, because they are generated by variations in magnetic
susceptibilities. Thus, fMRI studies are often limited by a reduced
BOLD sensitivity (BS) in inferior brain regions. Recently, a method has
been developed which allows for optimizing the BS in dropout regions by
specifically adjusting the slice tilt, the direction of the phase-encoding
(PE), and the z-shim moment. However, optimal imaging parameters
were only reported for the orbitofrontal cortex (OFC) and inferior
temporal lobes. The present study determines the optimal slice tilt, PE
direction, and z-shim moment at 3 Tand 1.5 T, otherwise using standard
fMRI acquisition parameters. Results are reported for all brain regions,
yielding a whole-brain atlas of optimal parameters. At both field
strengths, optimal parameters increase the BS by more than 60% in
many voxels in the OFC and by at least 30% in the other dropout
regions. BS gains are shown to be more widespread at 3 T, suggesting an
increased benefit from the dropout compensation at higher fields. Even
the mean BS of a large brain region, e.g., encompassing the medial OFC,
can be increased by more than 15%. The maps of optimal parameters
allow for assessing the feasibility and improving fMRI of brain regions
affected by susceptibility-induced BS losses.
© 2006 Elsevier Inc. All rights reserved.
Keywords: Signal loss; Susceptibility artifacts; EPI; fMRI; BOLD
sensitivity; z-shim
Introduction
Gradient-echo echo-planar imaging (GE EPI) is widely used for
functional magnetic resonance imaging (fMRI) of the brain
(Moonen and Bandettini, 2000). GE EPI is sensitive to microscopic
⁎ Corresponding author. Fax: +44 20 7813 1420.
E-mail address: [email protected] (N. Weiskopf).
Available online on ScienceDirect (www.sciencedirect.com).
1053-8119/$ - see front matter © 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2006.07.029
magnetic field alterations caused by blood oxygen level-dependent
(BOLD) susceptibility effects indicating neuronal activity (Logothetis, 2003). However, it is also sensitive to the macroscopic field
inhomogeneities caused by the differences of magnetic susceptibility of air and tissue which may result in local image distortions
and signal losses. While the image distortions may be corrected by
various on-line and off-line methods (e.g., Andersson et al., 2001;
Bowtell et al., 1994; Hutton et al., 2002; Jezzard and Balaban, 1995;
Sutton et al., 2004; Weiskopf et al., 2005b; Zaitsev et al., 2004),
signal dropouts have often compromised fMRI studies of the inferior
frontal, the medial temporal and the inferior temporal lobes (Devlin
et al., 2000; Merboldt et al., 2001; Ojemann et al., 1997).
Several methods have been devised for reducing susceptibilityrelated signal losses. Resistive shim coils have been used to
homogenize the magnetic field (Hsu and Glover, 2005). However,
strong magnetic field inhomogeneities can only be counterbalanced
by shimming a relatively small volume or relatively small spatial
orders. Signal losses may be reduced by using tailored radiofrequency (RF) pulses for excitation, opposing the dephasing effect
of the susceptibility related field gradients. However, quadratic
phase profile RF pulses reduce the signal-to-noise ratio (SNR) in
well-shimmed brain areas (Cho and Ro, 1992). Three-dimensional
tailored RF pulses are very long and require individual design of
the RF pulse for each subject before the fMRI experiment (Stenger
et al., 2000), a facility not yet readily available on standard
scanners. In general, in all directions smaller voxel sizes reduce the
signal loss but only at the expense of a reduced SNR in wellshimmed areas (Merboldt et al., 2000).
Z-shimming is a widely used compensation technique. A series
of images with different compensation gradient prepulses, counteracting the susceptibility-induced gradient in the slice selection
direction, is acquired and combined (e.g., Cordes et al., 2000;
Frahm et al., 1988; Ordidge et al., 1994). This method has been
extended to the phase encoding (PE) direction by Deichmann et al.
(2002). Although these methods can almost fully recover signal
losses, they significantly reduce the temporal resolution. Thus, they
compromise many fMRI studies requiring a high temporal
resolution, such as connectivity measurements (Friston et al.,
2003; Roebroeck et al., 2005) or event-related studies (Josephs et
494
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
al., 1997). A faster technique using optimal slice tilts (Deichmann
et al., 2003) and PE directions (De Panfilis and Schwarzbauer,
2005) combined with a moderate z-shim gradient prepulse has
been developed. The z-shimming gradient counteracts susceptibility-induced through-plane gradients whilst the slice tilt is used
to orient the slice so that the susceptibility-induced gradient in the
PE direction is small enough not to cause dropouts. The polarity of
the PE gradient can also be chosen optimally. These techniques,
combined together, have been shown to considerably enhance the
BOLD sensitivity (BS) in the orbitofrontal cortex (OFC;
Deichmann et al., 2003) and inferior temporal lobes (De Panfilis
and Schwarzbauer, 2005) whilst only slightly decreasing the BS in
well-shimmed areas. Since this approach only requires a single
acquisition, the high temporal resolution of standard EPI can be
retained.
We aimed to create 3D maps showing the scanning parameters
which optimize the BS for each voxel across the whole brain. In
contrast to previous works on the same subject which focused on
the OFC and the inferior temporal lobes, these maps allow for an
informed choice of the slice tilt, z-shim gradient moment, and PE
gradient polarity in all brain regions. To provide data, we scanned a
group of 5 subjects, varying systematically the slice tilt, the z-shim
gradient moment, and the PE gradient polarity. We scanned at two
commonly used field strengths of 3 T and 1.5 T using otherwise
standard fMRI imaging parameters. We estimated the local BS for
the group according to Deichmann et al. (2002) based on the
complex EPI raw data. We determined the maximal achievable BS
and respective optimal imaging parameters for each brain region
given that the z-shimming prepulse should not reduce the BS in
well-shimmed areas by more than 15%. To aid the decision about
which field strength should be used for scanning particular regions
of interest, we located regions where the BS achievable at 3 T
differed significantly from that at 1.5 T, when optimal parameter
sets have been chosen for each field strength.
Methods
Data acquisition and image processing
Five volunteers (1 female, 4 male, age 29–36 years) were
scanned in both a 1.5 T whole-body scanner (Magnetom Sonata,
Siemens Medical, Erlangen, Germany) and a 3 T head scanner
(Magnetom Allegra, Siemens Medical). The whole-body scanner
was operated with its standard body transmit and head receive coil
(single receive CP head array), the head scanner with its standard
birdcage transmit-receive coil. All volunteers gave written consent
as required by the local ethics committee. The participants were
scanned in supine position.
The manufacturer’s standard automatic 3D-shim procedure was
performed at the beginning of each experiment, to correct for first
and second order distortions in the static magnetic field in a given
adjust volume. The adjust volume corresponded to the volume
covered by the EPI imaging volume recorded with a slice tilt of β =
− 30° (for imaging details see below). The participants were
scanned with a single-shot EPI sequence which allowed for the free
choice of the moment of the z-shimming gradient prepulse (Mzcomp)
and the PE gradient polarity. The EPI imaging parameters at 1.5 T/
3 T were: 48 oblique transverse slices, slice thickness = 2 mm, gap
between slices = 1 mm, repetition time TR = 4.3/3.1 s, α = 90°, echo
time TE = 50/30 ms, bandwidth BW = 2298/3551 Hz/pixel, bandwidth in PE direction BWPE = 31.3/47.3 Hz/pixel, PE direction
anterior–posterior, field of view FOV = 192 × 192 mm2, matrix size
64 × 64. The z-shimming gradient moment was varied from
Mzcomp = −4 mT/m ms to + 4 mT/m ms (in 9 steps of 1 mT/m
ms), the slice tilt from β = − 45° to +30° (in 6 steps of 15°), and the
PE gradient was set to either positive or negative polarity, resulting
in 108 different parameter combinations. The slice tilt β is defined
as the angle of the rotation (pitch) about the x-axis/left–right
scanner axis measured from the axial plane ( = plane defined by the
left–right and the vertical scanner axis), with positive values
denoting tilts of the anterior edge of the slice towards the feet. A
positive PE gradient polarity corresponds to a positive PE
prewinder moment, i.e., the respective gradient points from the
posterior to the anterior part of the brain. For each parameter
combination a short time-series of 5 EPI volumes was acquired.
Further analysis was based on the last volume of the time-series to
exclude transitional T1 saturation effects. At 3 T, a posthoc
analysis of the BS in brain areas not affected by susceptibilityinduced field inhomogeneities showed that the data acquired at
TE = 30 ms was subject to a residual z-gradient moment of
0.6 mT/m ms. This effect may be due to either a slight offset of
the z-gradient caused by the shim coil or a miscalibration of the
slice refocusing gradient. Therefore, the effective z-shimming
gradient moments at 3 T were + 0.6 mT/m ms higher than the
nominal values, i.e., effectively ranging from − 3.4 mT/m ms to
+ 4.6 mT/m ms. In the following, all reported z-shimming
moments are the effective values. Although data were acquired
at a comprehensive range of z-shimming gradient moments, those
exceeding ± 2.5 mT/m ms were not included in the further analysis,
because according to the theory (Deichmann et al., 2003) they
cause significant BS loss (> 15%) in well-shimmed areas at a slice
thickness of 2 mm.
EPI magnitude images, echo time (TE) maps, and BS maps
were reconstructed from the complex k-space data using a
generalized reconstruction method based on the measured EPI kspace trajectory to minimize ghosting (Josephs et al., 2000). The
TE maps and BS maps were estimated according to Deichmann et
al. (2002). The BS was determined from the product of the image
intensity I and local TE: BS = TE × I. The image intensity was
extracted from the magnitude of the complex reconstructed image.
The local TE was determined from the local phase increment
across lines (in the PE direction) in the complex image: TE = TE0 +
(TE0 − t0)ΔΦ/π, with TE0 being the nominal echo time as entered
on the scanner interface, t0 being the time difference between RF
excitation and the start of the EPI readout, and ΔΦ being the phase
increment per line in the PE direction in the complex image. A
detailed derivation of the calculation of BS and TE can be found in
the Theory section and the Appendix C in Deichmann et al. (2002).
For anatomical reference and spatial normalization, a
high-resolution T1-weighted 3D image was acquired for each
participant at 1.5 T (3D MDEFT [Deichmann et al., 2004];
FOV = 256 × 224 mm2 or 256 × 240 mm2, matrix size 256 × 224 or
256 × 240, 176 partitions, slab thickness 176 mm, τ1 = 222.6 ms,
τ2 = 307.4 ms, TR = 12.24 ms, TE = 3.56 ms, α = 23o, BW = 106 Hz/
pixel). For each experiment, a field map using a double echo FLASH
sequence was recorded for distortion correction. The parameters at
1.5 T/3 T were: 64 oblique transverse slices, slice thickness = 2 mm,
gap between slices = 1 mm, TR = 1170/1020 ms, α = 90°, short
TE = 10 ms, long TE = 14.76/12.46 ms, BW = 260 Hz/pixel, PE
direction anterior–posterior, FOV = 192 × 192 mm2, matrix size
64 × 64, flow compensation). In one subject, only 48 slices were
sufficient to record the whole head field map (reduced TR = 874/
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
761 ms [1.5 T/3 T]). In one other experiment (at 1.5 T) the field map
was recorded with an increased matrix size of 128 × 128 and
FOV = 210 × 210 mm2 to avoid wrap over in the PE direction (slice
thickness = 2 mm, gap = 0.4 mm). Other imaging parameters were
adjusted accordingly.
Using the FieldMap toolbox (Hutton et al., 2002, 2004), field
maps were estimated from the phase difference between the images
acquired at the short and long TE. After spatial smoothing of the
field map (Gaussian kernel, FWHM = 6 mm), voxel displacements
in the EPI image were determined from the fieldmap and given
imaging parameters. Unwarping was performed by applying the
inverse displacement to the EPI magnitude images, TE maps, and
BS maps without a correction for the distortion-induced intensity
modulations. This correction was omitted, because it can reduce
the signal-to-noise ratio (SNR; Hutton et al., 2002). The T1weighted 3D image, field maps, and all images derived from EPI
(magnitude, TE, BS) were coregistered, resliced, and spatially
normalized using SPM2 (Ashburner and Friston, 1999; Friston et
al., 1995). The estimation of the parameters of the non-linear
spatial normalization was based on matching the individual T1weighted 3D image to the 152 brains template (Collins et al., 1994)
supplied by the Montreal Neurological Institute (MNI). For the
group analysis, the BS maps were masked to exclude non-brain
areas and areas not scanned in all subjects (by the inclusive
conjunction of the imaging volumes and the MNI template brain
mask) and spatially smoothed (Gaussian kernel, FWHM = 6 mm).
Maps: BS, BS gain, and optimal EPI parameters
A baseline BS map for each subject was determined from the
standard EPI acquisition (with slice tilt β = 0o, minimal effective
z-shimming moment Mzcomp = 0 mT/m ms at 1.5 T and Mzcomp =
− 0.4 mT/m*ms at 3 T, and a positive PE gradient polarity). To
allow for easier interpretation, each subject’s BS map was scaled to
100% for the mean BS value in well-shimmed brain areas of the
baseline BS map. The well-shimmed brain areas were defined as
the supplementary motor area (SMA), paracentral lobules,
precentral gyri, postcentral gyri, and the superior parietal gyri
identified according to the automated anatomical labeling toolbox
(AAL; Tzourio-Mazoyer et al., 2002).
For each voxel, we determined the mean BS across the
group for each acquisition parameter set. The optimal EPI
acquisition parameters (β, Mzcomp, PE gradient polarity) yielding
the highest mean BS were selected, excluding z-shimming
moments yielding a BS loss higher than 15% in well-shimmed
areas (|Mzcomp| < 2.5 mT/m ms). These parameters produced an
optimal BS map. The gain in BS using optimal acquisition
parameters was calculated by subtracting the baseline BS map
from the optimal BS map.
The BS maps are in relative rather than absolute units. To allow
for a comparison between field strengths, we assumed that the BS
should increase by approximately 20% in well-shimmed areas
going from 1.5 T to 3 T in an fMRI group study. This figure is a
conservative estimate based on studies comparing temporal SNR
(Triantafyllou et al., 2005) and BOLD CNR (Hoenig et al., 2005;
Turner et al., 2005) at 1.5 T and 3 T. All maps except for the
maximal BS maps were thresholded at a BS gain of 15%, to focus
on the areas where the dropout compensation significantly
improved the BS. For anatomical reference, all maps were
superimposed over the group averaged T1-weighted anatomical
image.
495
Region of interest analysis: optimal EPI parameters for a brain
region
In addition to selecting optimal parameters for each voxel,
we calculated the optimal parameter set for a selection of
extended brain regions which are frequently studied: namely,
the medial orbitofrontal cortex (mOFC) and rostral–ventral
anterior cingulate cortex (rACC) as the part of the OFC and
ACC suffering most from dropouts (orbital and medial orbital
part of the superior frontal gyri, olfactory cortex, gyrus rectus,
anterior cingulate with z < 0); the inferior temporal gyri;
temporal poles (superior and mid temporal pole); amygdala;
and hippo- and parahippocampal region. All regions of interest
were extracted from the AAL toolbox (Tzourio-Mazoyer et al.,
2002), except for the amygdala region which was defined by a
cytoarchitectonic probability map (p > 10%) described by
Amunts et al. (2005) using the SPM Anatomy toolbox
(Eickhoff et al., 2005). For each region we determined the
parameter set maximizing the mean BS across all voxels in the
region. To assess the inter-subject variation, the optimization
was performed both for the group mean and for each subject
individually.
Results
The head orientation of the subjects in each scanner differed
only minimally. The largest deviations were observed in the
rotation (pitch) about the left–right axis (the pitch was measured as
the angle between the scanner’s vertical axis and the AC–PC line
approximated by the x = 0, z = 0 line in MNI space). In the 3 T
scanner the pitch was on average − 14.9° ± 4.1° (mean ± standard
deviation), and in the 1.5 T scanner − 10.8° ± 4.0°.
Maps: BS, BS gain, and optimal EPI parameters
The BS and derived maps are shown in Fig. 1. Fig. 1a shows
that optimal imaging parameters increased the BS by more than
15% compared to the baseline BS in many brain regions inferior to
the anterior commissure (z = 0 mm). The maximal gains in BS
exceeded 60% at both field strengths. The gains in BS, however,
were more widespread at 3 T. Despite this increase in BS, inferior
frontal and temporal areas suffered a persistent dropout (e.g., the
U-shaped region at z = − 24 mm in Fig. 1b). As shown in Fig. 1c, in
the great majority of areas affected by signal dropout, higher BS
was observed for 3 T than 1.5 T. Only a small part of the left
inferior temporal gyrus (Fig. 1c, z = − 36 mm) showed a higher BS
at 1.5 T than at 3 T, exceeding the threshold of 15%. In four other
areas, higher BS at 1.5 T were observed as well, but all clusters
were smaller than 0.08 ml.
The distribution of optimal parameters followed a similar
coarse spatial pattern, being similar at 3 T (Fig. 2) and 1.5 T
(Fig. 3). Roughly speaking, negative z-shimming gradient
moments were optimal for dropout areas superior to z =
− 30 mm, positive moments for dropout areas inferior to z =
− 30 mm. For example, negative z-shimming gradient moments
were required for the OFC and rACC (z = −18 mm), and
posterior inferior temporal gyri (z = − 18 mm). Positive zshimming moments were optimal for parts of the temporal
poles, amygdalae, anterior part of the inferior temporal gyri,
hippo- and parahippocampus (z = − 30 mm). The z-shimming
moments were similar for both PE polarities.
496
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
Fig. 1. BOLD sensitivity (BS) at 1.5 T and 3 T using optimal EPI parameters. All maps are superimposed over axial slices of the group mean anatomical image (z-coordinates at the lower left). Areas not scanned in all
subjects and non-brain areas are excluded. (a) BS gain. Difference between the maximal BS using optimal EPI parameters and the BS achieved with standard EPI parameters (threshold >15%). The difference is
scaled to the mean BS (=100%) in a well-shimmed brain area using standard EPI parameters. (b) BS max. Maximal BS using optimal EPI parameters. Same scaling as in (a). (c) Difference between the BS at 3 T and
1.5 T. Same scaling as in (a), but the BS values at 3 T have been rescaled by 1.2, to adjust for the generally higher BS expected at 3 T.
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
Fig. 2. Optimal EPI parameters at 3 T. All maps are superimposed over axial slices of the group mean anatomical image (z-coordinates at the lower left). Areas not scanned in all subjects and non-brain areas are
excluded. Parameters are only shown for BOLD sensitivity (BS) gains exceeding 15%. (a) Optimal moment of the z-shimming prepulse (PP) given for each phase-encoding (PE) polarity. (b) Optimal slice tilt given
for each PE polarity. (c) Difference between the BS using the positive and the negative PE polarity. The difference is scaled to the mean BS (= 100%) in a well-shimmed brain area using standard EPI parameters.
497
498
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
Fig. 3. Optimal EPI parameters at 1.5 T. All maps are superimposed over axial slices of the group mean anatomical image (z-coordinates at the lower left). Areas not scanned in all subjects and non-brain areas are
excluded. Parameters are only shown for BOLD sensitivity (BS) gains exceeding 15%. (a) Optimal moment of the z-shimming prepulse (PP) given for each phase-encoding (PE) polarity. (b) Optimal slice tilt given
for each PE polarity. (c) Difference between the BS using the positive and the negative PE polarity. The difference is scaled to the mean BS (=100%) in a well-shimmed brain area using standard EPI parameters.
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
499
Fig. 4. Contribution of slice tilt and phase-encoding (PE) polarity to the gain in BOLD sensitivity (BS). Glass brains (maximum intensity projections) show the
additional gain in BS that can be achieved by simultaneously optimizing the slice orientation, the PE direction, and the z-shimming gradient moment, rather than
optimizing the z-shimming moment only with fixed slice orientation (axial) and fixed PE direction (both directions shown): at (a) 1.5 T and (b) 3 T. Only
differences greater than 20% are shown.
The optimal slice tilt showed a more heterogeneous spatial
pattern (Fig. 2b). For a positive PE, negative slice tilts were
optimal in the temporal pole (z = − 42 mm) and posterior OFC (z =
− 18 mm), whereas positive slice tilts were optimal in the anterior
OFC and inferior temporal gyri (z = − 18 mm). In most regions, the
sign of the slice tilt changed with the PE polarity.
Figs. 2c and 3c show that positive PE polarities improved BS in
the posterior OFC and caudate nuclei (z = − 6 mm), whereas
negative PE polarities improved the BS in the anterior OFC (z =
− 18 mm), parts of the fusiform gyri (around z = −42 mm), temporal
poles (z = − 44 mm), and inferior and middle temporal gyri (z =
− 30 mm).
Fig. 4 shows the additional BS gain that can be achieved by
simultaneously optimizing the slice orientation, the PE direction,
and the z-shimming prepulse moment, rather than optimizing the
latter only with fixed slice orientation and fixed PE direction. The
shown values correspond to the difference of the maximally
achievable BS when optimizing all parameters, minus the
maximally achievable BS when optimizing the z-shimming moment
with a fixed slice tilt of 0o and a positive/negative PE polarity.
Region of interest analysis: optimal EPI parameters for a brain
region
Table 1
Optimal EPI parameters for regions of interest at 3 T
Table 2
Optimal EPI parameters for regions of interest at 1.5 T
Region of interest
Tilt PE
PP
[mT/m*ms] [deg]
Optimal
Standard
mean % BS mean % BS
mOFC+rACC
Inferior temporal
lobes
Temporal poles
Amygdala
Hippocampus +
Parahippocampus
−1.4
−0.4
− 45
+30
neg.
neg.
99.2
78.1
83.5
75.1
+0.6
+0.6
+0.6
+30
− 45
− 45
neg. 106.0
pos. 120.9
pos. 121.9
89.5
107.3
109.7
PP = z-shimming prepulse gradient moment; PE = phase-encoding polarity;
BS = BOLD sensitivity; mOFC+rACC = medial orbitofrontal cortex and
rostral–ventral anterior cingulate cortex with z < 0 mm (MNI).
Tables 1 and 2 show that in all regions of interest the mean BS
was improved by an optimal parameter choice. Increases in the BS
were in general smaller at 1.5 T (Table 2) than at 3 T (Table 1). The
most consistent increase in mean BS was observed for the mOFC+
rACC, exceeding 15% at 3 T and 1.5 T. In Fig. 5a the histograms of
the BS distribution of voxels in the mOFC+rACC for the
optimized and standard EPI are plotted. They show that the
increase in mean BS is achieved both by increasing the number of
voxels with BS > 120% and also decreasing the number of voxels
with BS < 75%. Therefore, the increase in mean BS is not only a
result of increasing the BS in voxels with a high baseline BS, but
also due to a significant reduction in the number of voxels with a
low BS. At 3 T, prominent BS increases were found additionally in
the temporal poles (16%, Table 1) and the amygdalae (14%, Table
1). The optimal z-shimming moments for regions did not exceed
the range from + 1 mT/m ms to − 1.4 mT/m ms, although for single
voxels the optimal values were occasionally larger (e.g., Fig. 2a,
Region of interest
Tilt PE
PP
[mT/m*ms] [deg]
Optimal
Standard
mean % BS mean % BS
mOFC+rACC
Inferior temporal
lobes
Temporal poles
Amygdala
Hippocampus +
Parahippocampus
− 1.0
− 1.0
−45
+30
neg.
pos.
91.6
84.1
76.4
80.4
+1.0
+0.0
+0.0
+30
−45
−4 5
neg. 105.8
pos. 120.3
pos. 122.7
98.5
115.0
118.3
PP = z-shimming prepulse gradient moment; PE = phase-encoding polarity;
BS = BOLD sensitivity; mOFC+rACC = medial orbitofrontal cortex and
rostral–ventral anterior cingulate cortex with z < 0 mm (MNI).
500
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
Discussion
This study provides guidelines for maximizing the BOLD
sensitivity (BS) of echo-planar imaging (EPI) in brain areas
affected by signal dropout at 3 T and 1.5 T. It expands upon a
previously established dropout compensation technique (De
Panfilis and Schwarzbauer, 2005; Deichmann et al., 2003) focusing
on the optimal choice of the slice tilt, phase-encoding (PE)
direction, and z-shimming gradient moment and using otherwise
conventional basic EPI parameters (echo time TE = 30/50 ms at
3 T/1.5 T; voxel size 3 × 3 × 2 mm3). The appropriate choice of slice
tilt and PE polarity primarily reduces BS loss due to susceptibilityinduced gradients in the PE direction. Moderate z-shimming
reduces signal loss due to through-plane gradients.
BOLD sensitivity and optimal EPI parameters
Fig. 5. EPI parameter optimization for regions of interest at 3 T: BOLD
sensitivity (BS) histograms. The distribution of BS of the voxels inside (a)
the medial orbitofrontal cortex and rostral–ventral anterior cingulate cortex,
and (b) the temporal poles is plotted for the optimal parameters (solid black
line) and standard parameters (dashed dotted gray line). For visual guidance
the vertical double line marks a BS of 85%, indicating a significantly
reduced BS. The BS in each voxel is scaled to the mean BS (=100%) in a
well-shimmed brain area using standard EPI parameters.
z = − 18 mm). The BS gain was only slightly increased (by less than
1.1%) for any of the ROIs when optimizing the parameters for each
subject individually instead of using a single optimal parameter set
for the whole group.
Optimizing parameters for a given ROI also affected the BS in
the rest of the brain. To illustrate this effect, Fig. 6 shows
maximum intensity projections (MIPs) of BS gains and losses
across the whole brain when using the optimal parameters for the
mOFC+rACC, inferior temporal lobes, and temporal poles at 3 T.
The BS losses and gains exceeded 20% in many voxels across the
brain. For example, when using the optimal parameters for
investigation of the temporal poles, the BS in these areas increased
(Fig. 6c, left), but there were BS losses in the posterior OFC and
rACC (Fig. 6c, right). Fig. 7 presents single subject BS maps in
different brain areas, using the standard EPI sequence (left), and a
sequence with parameters that were optimized for the respective
area (right).
Since the direction and magnitude of the susceptibility-induced
gradient varies across the brain, the optimal EPI parameters are
location dependent. We have computed maps of optimal parameters
for each voxel in the brain (Figs. 2 and 3) and show that, in single
voxels (Fig. 1a, z = − 18 mm), an optimal choice of parameters can
increase the BS by more than 60% compared to conventional EPI
(100% = baseline EPI BS in well-shimmed areas). However, full
recovery of the BS is not possible in all regions, because we are
limited to the application of a constant moderate z-shimming
gradient moment so that BS loss is restricted in well-shimmed areas
(< 15% of baseline EPI BS) and high temporal resolution is retained
(Deichmann et al., 2003). Also, the compensation technique does
not counteract susceptibility-induced gradients in the EPI readout
(RO) direction (Weiskopf et al., 2005a). Regardless of the parameter
choice, areas of large residual BS loss persist, mainly confined to the
OFC and inferior temporal lobes (Fig. 1b, z = −24 to −30 mm).
At 3 T and 1.5 T we observed similar patterns of optimal
parameters (Figs. 2 and 3; Tables 1 and 2). This is not surprising,
because we expect the spatial distribution of the field inhomogeneities to be similar and to differ mainly in its magnitude (scaled by
the static magnetic field).
Approximately speaking, negative z-shimming gradient moments were optimal for dropout areas superior to z = − 30 mm, and
positive moments for dropout areas inferior to z = −30 mm (Figs.
2a and 3a). For example, the temporal poles, amygdalae, hippoand parahippocampus all benefit from positive z-shimming
moments, and the OFC from a negative one. Positive PE polarities
improved BS in the posterior OFC and caudate nuclei, whereas
negative PE polarities improved BS in the anterior OFC, and parts
of the inferior temporal lobes (Figs. 2c and 3c). The optimal slice
tilt parameter showed a more heterogeneous spatial pattern (Figs.
2b and 3b). For the positive PE polarity, negative slice tilts were
optimal in the temporal pole and posterior OFC, and positive slice
tilts in the anterior OFC and inferior temporal gyri. In most brain
regions, the sign of the optimal slice tilt reversed together with the
PE polarity. The data support the assumption that simultaneous
optimization of z-shimming moment, slice tilt, and PE polarity
leads to an improved recovery of BS losses, compared to the
optimization of only one parameter. As an example, it is shown that
only optimizing the z-shimming moment does not yield the same
BS gains as optimizing all three parameters (Fig. 4).
Since many investigators using fMRI are interested in brain
regions extending over larger areas, we determined parameter sets
optimizing the mean BS for a variety of regions of interest (ROI;
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
501
Fig. 6. BOLD sensitivity (BS) gains and losses when applying parameter sets optimized for (a) medial orbitofrontal cortex and rostral–ventral ACC (mOFC+
rACC), (b) inferior temporal lobes, (c) temporal poles. Left column of glass brains (maximum intensity projections) shows the gains in BS using the optimized
parameters compared to the standard EPI scan, while the right column shows the accompanying losses. Only differences greater than 20% are shown.
see Tables 1 and 2): the medial OFC (mOFC) and rostral–ventral
anterior cingulate cortex (rACC); the inferior temporal lobes;
temporal poles; amygdalae; and hippo- and parahippocampal
region. The most prominent mean BS recovery (approximately
15%) was achieved in the mOFC+rACC, reflecting the extended
area of dropout in this region. At 3 T, a considerable BS gain was
also achieved in the temporal poles (16%) and the amygdalae
(14%). The smaller increase of the mean BS in the other ROIs was
most likely caused by partial volume effects. Nevertheless, the gain
may be very significant in single voxels or parts of these areas and
sensitivity loss is restricted to a worst case of 15% in well-shimmed
areas. Obviously, the optimization must be redone if one is
interested in a smaller subpart of the ROIs or altered ROIs.
limit of about 20% BS increase in well-shimmed areas and rescale
the BS maps accordingly. Using such recalibrated BS maps, a
direct comparison between the two field strengths is possible and
suggests that 3 T provides a higher BS than 1.5 T in most brain
areas (Fig. 1c). Thus, fMRI at higher fields in areas affected by
susceptibility-induced gradients benefits more from dropout
compensation techniques and so any comparison of BS between
different field strengths should consider these techniques to avoid a
potential bias. However, we advise caution. We did not directly
measure and compare BS, but linked one set of results to the other
on the basis of literature values. Moreover, the 3 T scanner was
equipped with a head gradient system enabling a very rapid EPI
readout.
Comparison of 3 T and 1.5 T
Limitations and considerations
In general, the increases in BS using optimal EPI parameters
were more substantial at 3 T than at 1.5 T, as can be seen, for
example, in the temporal lobes (Fig. 1a, z = − 36 mm) or from the
larger BS increases in the selected regions of interest (Tables 1 and
2). The higher effectiveness of the dropout compensation at 3 T
supports the notion that 3 T may retain its generally higher BS
compared to 1.5 T even in areas affected by susceptibility-induced
gradients. Based on literature values for typical increases in BS
going from 1.5 T to 3 T (Hoenig et al., 2005; Triantafyllou et al.,
2005; Turner et al., 2005), one can conservatively estimate a lower
We took particular care to accurately estimate the BS from
single-shot image data. All images were corrected for geometric
distortions, which can otherwise simulate signal loss by shifting
signal from one area into other areas. Distortion-induced intensity
modulations were not corrected, because the correction can reduce
the signal-to-noise ratio (SNR; Hutton et al., 2002). However, in
GE EPI these intensity modulations are expected to be significantly
smaller than the signal loss caused by the intravoxel dephasing.
Importantly, our BS estimate was based not only on the measured
signal amplitude but also took local echo time shifts (Deichmann et
502
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
Fig. 7. Exemplary BOLD sensitivity (BS) maps of standard EPI (left) and
EPI optimized (right) for scanning the (a) medial orbitofrontal cortex and
rostral–ventral anterior cingulate cortex (mOFC+rACC), (b) inferior
temporal lobes, and (c) temporal poles at 3 T. Smoothed single subject BS
estimates are presented as axial slices (z-coordinates indicated in the upper
right corner). Areas not scanned for all slice tilts are excluded.
al., 2002) caused by susceptibility-induced gradients in the PE
direction into account.
Since we performed neither an fMRI experiment nor acquisition
of a time-series of images to estimate BS and temporal noise
(including physiological noise components; e.g., Krüger et al.,
2001), the reported BS values should be interpreted with particular
care regarding possible temporal noise variations. For example,
regions close to contrast edges may suffer from motion artifacts and
regions close to larger vessels from cardiac artifacts. Scanner
instability and the Nyquist ghosting in EPI (Josephs et al., 2006) can
be another source of temporal noise. In these cases the BS based only
on thermal noise may be overestimated. The BS values were
measured relative to the BS in a well-shimmed area in a conventional
EPI. Since the reference value is the average BS in a large ROI,
signals from different types of brain tissue (gray, white matter,
cerebrospinal fluid) are mixed. A spatial variation in their relative
density can therefore be misinterpreted as a different BS. Similarly,
variations in the RF coil sensitivity and B1 field inhomogeneities can
influence the estimation of the local BS. A non-uniformity of
calculated BS values across the brain due to different local coil
sensitivities or different relative tissue densities may complicate the
comparison of BS in different brain areas and bias the estimation of
optimal parameters for extended ROIs. However, the variations
would not affect the estimation of the optimal parameter maps,
because they are constant within a single voxel for all parameters.
This study was based on a group of 5 subjects instead on a
single subject, to provide a more representative estimate of the BS.
We regarded it as reasonable to pool multiple subjects, because the
relatively similar head orientation of the subjects in the scanners
(standard deviation of head tilt <5°) will generate similar field
inhomogeneities. Any small variation in pitch is expected to yield
similar BS changes as the corresponding change in the slice tilt
(Deichmann et al., 2003). The similar head orientation also allows
for a fixed spatial coordinate system based on the scanner to define
the slice tilts, instead of using different individual head coordinate
systems. Obviously, our results pertain only to this (common)
positioning. The pooling of multiple subjects is further supported
by the result that the mean BS gains in all ROIs did not
significantly increase (by less than 1.1%) when the parameters
were optimized for each subject individually instead of using a
single optimal parameter set for the whole group.
The group analysis required spatial normalization and smoothing of the data. As in all group analyses, both the reduced effective
spatial resolution and the inter-individual differences in brain
anatomy make it difficult to locate and delineate small brain
structures. This may impair extracting optimal parameters for small
regions or exact delineation and separation of different regions.
Similarly, since the regions of interest investigated in this study are
mainly based on the single-subject AAL atlas (Tzourio-Mazoyer et
al., 2002), the derived parameters are subject to mismatch between
individual anatomies and so should be regarded only as
approximate guidelines. More reliable estimates with error margins
may be possible with the increased availability of cytoarchitectonic
probability maps, such as the one used to define the amygdala
(Amunts et al., 2005).
Only a subset of possible EPI acquisition parameters was studied.
Firstly, only axial and oblique axial slice orientations with left–right
readout gradient direction were considered. This is reasonable,
because these are the most frequently used for fMRI and allow for
the fastest acquisition, where peripheral nerve stimulation is a limitation. Secondly, only moderate z-shimming moments were allowed
to avoid widespread signal loss in well-shimmed areas. In fact,
because of technical problems the maximal negative z-shimming
moment was slightly larger at 3 T than at 1.5 T which may have
contributed to the larger BS gains we observed at 3 T in the OFC.
Thirdly, slice tilts were limited to less than 45°, because special care
must be taken with large slice tilts, in particular positive ones
(positive values denote tilts of the anterior edge of the slice towards
the feet measured from the axial plane). In this case the slice may
intersect with the oral cavity and nose, causing wrap over, chemical
shift artifacts (from unsuppressed fatty tissue signals) and larger
Nyquist ghost. Therefore, we would recommend using negative
slice tilts if possible. Another solution may be to saturate the signal
from these areas or to use 3D spatially selective RF excitation pulses.
The presented dropout compensation is particularly useful for
recovering BS where two or more regions are affected by similar
susceptibility-induced gradients. If two different regions require
opposite z-shimming moments, the optimal parameter set for one
region can significantly reduce the BS (by much more than 15%) for
the other brain region (see Fig. 6 and Deichmann et al., 2003). Our
whole-brain maps help to identify such situations. They can even
help to overcome the problem by calibrating slice-dependent
z-shimming (Weiskopf and Deichmann, 2006). Slice-dependent zshimming uses the optimal gradient moment for each region separately (Cordes et al., 2000), at least where they can be segregated in
the slice select direction.
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
General applicability: scanner and sequence parameters
Since our imaging data were acquired on two particular MRI
scanner setups, it should be checked whether the specific results
and guidelines are applicable for a different scanner setup. For
example both scanners were equipped with second order shim coils
and the manufacturer’s automatic 3D shim was performed prior to
the EPI acquisition. If only first order shims are available, the BS
may be reduced. The 3 T scanner was operated with a head
gradient system enabling a very rapid EPI readout.
The mean head position in the scanner should be measured for a
group of subjects, in order to ensure that the tilt angles are
comparable. Any differences in head tilts are most likely about the
x-axis (pitch). A small difference in pitch may be approximated by
the corresponding slice tilt (Deichmann et al., 2003) and the
optimal slice tilts may be adjusted accordingly.
The EPI sequence parameters should be chosen to be as similar
as possible. For example, longer echo times require proportionally
increased z-shimming gradient moments (Deichmann et al., 2003).
However, despite the application of an increased z-shimming
gradient moment BS loss may still be increased for longer echo
times due to in-plane susceptibility-induced gradients.
If an increased slice thickness is used, the gradient moment
must be reduced reciprocally to avoid excessive BS losses due to
dephasing in well-shimmed brain areas (Deichmann et al., 2003).
However, we would not advise to use thicker slices because the
reduced z-shimming gradient moment and larger slice thickness
will both yield significantly larger BS losses in areas affected by
field inhomogeneities.
A change in the EPI readout time (echo spacing) and a change
in the in-plane resolution will both affect the BS loss due to inplane susceptibility-induced gradients. Increasing the in-plane
resolution (extending k-space coverage) will reduce both dropouts
due to susceptibility-induced gradients in the readout (Weiskopf et
al., 2005a) and PE direction (Deichmann et al., 2002), if the
readout time and other sequence parameters are kept constant. On
the other hand, an increased resolution will generally reduce the
signal-to-noise ratio (SNR) and thus the BS in well-shimmed areas
(Merboldt et al., 2000).
However, often the readout time is increased with a higher spatial
resolution. An increased readout time will primarily affect dropouts
due to susceptibility-induced gradients in the PE direction. The
increased readout time will increase any shifts to longer effective
echo times and exacerbate dropouts due to the shift of the echo
outside the acquisition window (Deichmann et al., 2002).
Paradoxically, a longer readout time may reduce the BS losses that
are caused by shifts to smaller effective echo times, because the
readout starts earlier after the RF excitation. However, this type of
BS loss occurs only for significantly higher field distortions in the
PE direction (Deichmann et al., 2003) and is rare. A longer readout
time will also yield larger geometric distortions (Hutton et al., 2002).
The readout time may also be increased when using a lower receiver
bandwidth, causing the same problems. However, lower bandwidths
increase the SNR and BS (Zou et al., 2005).
For other principal imaging orientations (i.e., sagittal or
coronal) or different PE directions (e.g., left–right), it would be
very difficult if not impossible to extrapolate optimal parameters
from the results presented here. However, other studies assessed
effects of different principal imaging orientations on BS, for
example for the amygdala (Chen et al., 2003; Robinson et al.,
2004; Merboldt et al., 2001). One should be aware that geometric
503
distortions depend on the PE direction (Bowtell et al., 1994;
Weiskopf et al., 2005b) and may compromise direct comparisons
between studies acquired with different PE directions, unless they
are corrected for.
Conclusion
This paper provides guidelines for optimizing echo-planar
imaging (EPI) for functional imaging studies in regions affected by
susceptibility-induced BOLD sensitivity (BS) losses at two field
strengths (1.5 T and 3 T). Whole-brain maps show the slice tilt,
phase-encoding (PE) polarity, and z-shimming gradient moment
that maximize the BS in each voxel. Furthermore, optimal
parameters are reported for selected regions of interest extending
across larger areas (e.g., the medial orbitofrontal cortex). The maps
can be used to assess the feasibility and to optimize fMRI of
regions affected by BS loss. The maps can also help in choosing
which field strength may be optimal.
Acknowledgments
This study was supported by the Wellcome Trust. We thank S.B.
Eickhoff for the help with extracting cytoarchitectonic probability
maps from the SPM Anatomy toolbox and J. O’Doherty for
providing his expertise on neuroanatomy.
References
Amunts, K., Kedo, O., Kindler, M., Pieperhoff, P., Mohlberg, H., Shah, N.J.,
Habel, U., Schneider, F., Zilles, K., 2005. Cytoarchitectonic mapping of
the human amygdala, hippocampal region and entorhinal cortex: intersubject variability and probability maps. Anat. Embryol. 210, 343–352.
Andersson, J.L.R., Hutton, C., Ashburner, J., Turner, R., Friston, K., 2001.
Modeling geometric deformations in EPI time series. NeuroImage 13,
903–919.
Ashburner, J., Friston, K.J., 1999. Nonlinear spatial normalization using
basis functions. Hum. Brain Mapp. 7, 254–266.
Bowtell, R.W., McIntyre, D.J.O., Commandre, M.-J.,Glover, P.M., Mansfield,
P., 1994. Correction of geometric distortion in echo planar images. Proc.
SMR, 2nd Annual Meeting, Society of Magnetic Resonance, Berkeley,
CA, USA, p. 411.
Chen, N.K., Dickey, C.C., Yoo, S.S., Guttmann, C.R., Panych, L.P., 2003.
Selection of voxel size and slice orientation for fMRI in the presence of
susceptibility field gradients: application to imaging of the amygdala.
NeuroImage 19, 817–825.
Cho, Z.H., Ro, Y.M., 1992. Reduction of susceptibility artifact in gradientecho imaging. Magn. Reson. Med. 23, 200.
Collins, D.L., Neelin, P., Peters, T.M., Evans, A.C., 1994. Automatic 3D
intersubject registration of MR volumetric data in standardized Talairach
space. J. Comput. Assist. Tomogr. 18, 192–205.
Cordes, D., Turski, P.A., Sorenson, J.A., 2000. Compensation of
susceptibility-induced signal loss in echo-planar imaging for functional
applications. Magn. Reson. Imaging 18, 1055–1068.
Deichmann, R., Josephs, O., Hutton, C., Corfield, D.R., Turner, R., 2002.
Compensation of susceptibility-induced BOLD sensitivity losses in
echo-planar fMRI imaging. NeuroImage 15, 120–135.
Deichmann, R., Gottfried, J.A., Hutton, C., Turner, R., 2003. Optimized EPI
for fMRI studies of the orbitofrontal cortex. NeuroImage 19, 430–441.
Deichmann, R., Schwarzbauer, C., Turner, R., 2004. Optimisation of the 3D
MDEFT sequence for anatomical brain imaging: technical implications
at 1.5 and 3 T. NeuroImage 21, 757–767.
De Panfilis, C., Schwarzbauer, C., 2005. Positive or negative blips? The
effect of phase encoding scheme on susceptibility-induced signal losses
in EPI. NeuroImage 25, 112–121.
504
N. Weiskopf et al. / NeuroImage 33 (2006) 493–504
Devlin, J.T., Russell, R.P., Davis, M.H., Price, C.J., Wilson, J., Moss, H.E.,
Matthews, P.M., Tyler, L.K., 2000. Susceptibility-induced loss of
signal: comparing PET and fMRI on a semantic task. NeuroImage 11,
589–600.
Eickhoff, S.B., Stephan, K.E., Mohlberg, H., Grefkes, C., Fink, G.R.,
Amunts, K., Zilles, K., 2005. A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data. NeuroImage
25, 1325–1335.
Frahm, J., Merboldt, K.D., Hänicke, W., 1988. Direct FLASH MR imaging
of magnetic field inhomogeneities by gradient compensation. Magn.
Reson. Med. 6, 480.
Friston, K.J., Ashburner, J., Frith, C.D., Poline, J.-B., Heather, J.D.,
Williams, S.C., Frackowiack, R.S.J., 1995. Spatial registration and
normalization of images. Hum. Brain Mapp. 2, 165–189.
Friston, K.J., Harrison, L., Penny, W., 2003. Dynamic causal modelling.
NeuroImage 19, 1273–1302.
Hoenig, K., Kuhl, C.K., Scheef, L., 2005. Functional 3.0-T MR assessment
of higher cognitive function: are there advantages over 1, 5-T imaging?
Radiology 234, 860–868.
Hsu, J.-J., Glover, G.H., 2005. Mitigation of susceptibility-induced signal
loss in neuroimaging using localized shim coils. Magn. Reson. Med. 53,
243–248.
Hutton, C., Bork, A., Josephs, O., Deichmann, R., Ashburner, J., Turner, R.,
2002. Image distortion correction in fMRI: a quantitative evaluation.
NeuroImage 16, 217–240.
Hutton, C., Deichmann, R., Turner, R., Andersson, J.L.R., 2004. Combined
correction for geometric distortion and its interaction with head motion
in fMRI. Proceedings of ISMRM 12, Kyoto, Japan, p. 1084.
Jezzard, P., Balaban, R.S., 1995. Correction for geometric distortion in
echo planar images from B0 field variations. Magn. Reson. Med. 34,
65–73.
Josephs, O., Turner, R., Friston, K., 1997. Event-related fMRI. Hum. Brain
Mapp. 5, 243–248.
Josephs, O., Deichmann, R., Turner, R., 2000. Trajectory measurement and
generalised reconstruction in rectilinear EPI. Proceedings of ISMRM 8,
Denver, Colorado, USA, p. 1517.
Josephs, O., Weiskopf, N., Deichmann, R., 2006. Correction of EPI Nyquist
ghosting caused by the non-uniform frequency response of the MRI
receiver chain. Proceedings of ISMRM 14, Seattle, Washington, USA.
Krüger, G., Kastrup, A., Glover, G.H., 2001. Neuroimaging at 1.5 T and
3.0 T: comparison of oxygenation-sensitive magnetic resonance
imaging. Magn. Reson. Med. 45, 595–604.
Logothetis, N.K., 2003. MR-imaging in the non-human primate: studies of
function and of dynamic connectivity. Curr. Opin. Neurobiol. 13,
630–642.
Merboldt, K.D., Finsterbusch, J., Frahm, J., 2000. Reducing inhomogeneity
artifacts in functional MRI of human brain activation-thin sections vs.
gradient compensation. J. Magn. Reson. 145, 184–191.
Merboldt, K.D., Fransson, P., Bruhn, H., Frahm, J., 2001. Functional MRI of
the human amygdala? NeuroImage 14, 253–257.
Moonen, C.T.W., Bandettini, P.A., 2000. Functional MRI. Springer, Berlin.
Ojemann, J., Akbudak, E., Snyder, A.Z., McKinstry, R.C., Raichle, M.E.,
Conturo, T.E., 1997. Anatomic localization and quantitative analysis of
gradient refocused echo-planar fMRI susceptibility artifacts. NeuroImage 6, 156–167.
Ordidge, R.J., Gorell, J.M., Deniau, J.C., Knight, R.A., Helpern, J.A., 1994.
Assessment of relative brain iron concentrations using T2-weighted and
T2*-weighted MRI at 3 Tesla. Magn. Reson. Med. 32, 335–341.
Robinson, S., Windischberger, C., Rauscher, A., Moser, E., 2004. Optimized
3 T EPI of the amygdalae. NeuroImage 22, 203–210.
Roebroeck, A., Formisano, E., Goebel, R., 2005. Mapping directed
influence over the brain using Granger causality and fMRI. NeuroImage
25, 230–242.
Stenger, V.A., Boada, F.E., Noll, D.C., 2000. Three-dimensional tailored RF
pulses for the reduction of susceptibility artifacts in T2*-weighted
functional MRI. Magn. Reson. Med. 44, 525–531.
Sutton, B.P., Noll, D.C., Fessler, J.A., 2004. Dynamic field map estimation
using a spiral-in/spiral-out acquisition. Magn. Reson. Med. 51,
1194–1204.
Triantafyllou, C., Hoge, R.D., Krueger, G., Wiggins, C.J., Potthast, A.,
Wiggins, G.C., Wald, L.L., 2005. Comparison of physiological noise at
1.5 T, 3 T and 7 T and optimization of fMRI acquisition parameters.
NeuroImage 26, 243–250.
Turner, R., Mechelli, A., Noppeney, U., Price, C., Glensman, J., Rombouts,
S., Veltman, D., 2005. A comparison of the effectiveness of 3 T and 1.5 T
MRI scanners in a multi-subject fMRI study of visual memory. Abstract
at 11th Annual Meeting of the OHBM, Toronto, Ontario, Canada.
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard,
O., Delcroix, N., Mazoyer, B., Joliot, M., 2002. Automated anatomical
labeling of activations in SPM using a macroscopic anatomical
parcellation of the MNI MRI single- subject brain. NeuroImage 15,
273–289.
Weiskopf, N., Deichmann, R., 2006. Automated variable z-shim for fMRI:
user-independent reduction of signal dropout in the orbitofrontal cortex
and temporal lobe. Abstract at 12th Annual Meeting of the OHBM,
Florence, Italy.
Weiskopf, N., Hutton, C., Josephs, O., Turner, R., Deichmann, R., 2005a.
Signal dropouts in EPI caused by susceptibility gradients in the
frequency encoding direction: theory and experimental validation.
Abstract at 11th Annual Meeting of the OHBM, Toronto, Ontario,
Canada.
Weiskopf, N., Klose, U., Birbaumer, N., Mathiak, K., 2005b. Singleshot compensation of image distortions and BOLD contrast
optimization using multi-echo EPI for real-time fMRI. NeuroImage
24, 1068–1079.
Zaitsev, M., Hennig, J., Speck, O., 2004. Point spread function mapping
with parallel imaging techniques and high acceleration factors: fast,
robust, and flexible method for echo-planar imaging distortion
correction. Magn. Reson. Med. 52, 1156–1166.
Zou, P., Hutchins, S.B., Dutkiewicz, R.M., Li, C.S., Ogg, R.J., 2005. Effects
of EPI readout bandwidth on measured activation map and BOLD
response in fMRI experiments. NeuroImage 27, 15–25.